Railway turnout safety control system and method based on radar induction

Abstract

The application relates to the technical field of railway safety control, and particularly discloses a railway turnout safety control system and method based on radar induction, which obtains echo signals of a railway turnout throat area through a millimeter wave radar, forms a distance speed spectrum matrix, selects a stable scattering reference unit in a safety protection area, continuously extracts a phase angle of the reference unit and calculates a signal disturbance index. When a vehicle enters the safety protection area, the vehicle drives conductive dust disturbance, so that the phase fluctuation of the reference unit is enhanced; the index is compared with a dynamic baseline threshold value, an intrusion disturbance mark is generated, and a turnout locking control signal is output.

Description

Technical Field

[0001] This invention relates to the field of railway safety control technology, specifically to a railway turnout safety control system and method based on radar sensing. Background Technology

[0002] In railway systems located in steel mills and other industrial areas with high levels of conductive dust, conductive solid particles such as iron powder and steel slag are constantly suspended in the throat area of ​​turnouts, severely impacting millimeter-wave radar detection. When these suspended conductive particles are irradiated by electromagnetic waves, they generate induced currents, causing energy dissipation. Simultaneously, the particles act as secondary radiation sources, further scattering electromagnetic waves, creating a double attenuation effect and significantly reducing the effective detection range of radar for vehicle targets.

[0003] Current radar detection solutions generally rely on echo signal strength to determine the presence of a target. In the scenario described above, the signal attenuation caused by conductive dust and the absence of vehicles in the area cannot be effectively distinguished solely by signal amplitude. The radar equipment appears normal in appearance and operation, but its actual detection capability has significantly degraded, and operators are unaware of the protection failure. When a vehicle approaches or intrudes into the switch area, the safety control system fails to trigger the switch locking action properly, easily leading to safety accidents such as switch derailment and train derailment. Summary of the Invention

[0004] The purpose of this invention is to provide a radar-sensing-based railway turnout safety control system and method to solve the above-mentioned technical problems.

[0005] The objective of this invention can be achieved through the following technical solutions: A radar-sensing-based railway turnout safety control method includes the following steps: Control the millimeter-wave radar to transmit electromagnetic wave signals toward the throat area of ​​the railway turnout and receive the echo signals; The echo signal is processed to generate digital difference frequency data, and then a fast Fourier transform is performed on multiple consecutive frames of digital difference frequency data to obtain the range-velocity spectrum matrix. A reference cell is selected in the range-velocity spectrum matrix, the phase angle of the reference cell is extracted frame by frame and a phase angle sequence is established, and the signal disturbance index is calculated based on the phase angle sequence. Read the currently stored baseline value, which includes the baseline mean and baseline standard deviation, and determine the current discrimination threshold based on the baseline value; The signal disturbance index is compared with the discrimination threshold. When the discrimination condition is met, an intrusion disturbance flag is generated. Based on the intrusion disturbance flag, the turnout locking control signal is output.

[0006] As a further aspect of the present invention, the process of obtaining the range-velocity spectrum matrix is ​​as follows: The linear frequency modulated continuous wave signal generated by the voltage-controlled oscillator of the millimeter-wave radar is used as the transmission signal and is divided into a transmission branch and a local oscillator branch by a power divider; The echo signal received by the receiving antenna is amplified by low noise and then mixed with the local oscillator branch to obtain a difference frequency signal containing the distance information of the scatterer; After the difference frequency signal is filtered to remove the DC component and high-frequency noise, it is converted into a digital difference frequency sequence by an analog-to-digital converter. A one-dimensional fast Fourier transform is performed on the digital difference frequency sequence within each frequency modulation cycle to obtain the range spectrum. The range spectra corresponding to multiple consecutive frequency modulation cycles are arranged into a matrix in time order, and then a one-dimensional fast Fourier transform is performed along the slow time dimension to form a range velocity spectrum matrix, where the frequency shift in the slow time dimension corresponds to the radial velocity of the scatterer.

[0007] As a further aspect of the present invention, the process of obtaining the phase angle sequence is as follows: In the range-velocity spectrum matrix, determine the column index A of the zero velocity channel. Each range cell in the zero velocity channel corresponds to the part of the scatterer where the radial velocity is zero. Search for range cells whose amplitude value is continuously higher than the noise threshold within the range of the two columns to the left and right of column index A. The searched adjacent distance cells are combined into a fixed scattering distance segment. The distance cells located within the safety protection area are selected as candidate cells, and the distance cell with the largest average amplitude is selected as the reference cell from the candidate cells. The complex reflection coefficients of the reference unit are obtained frame by frame, and the real and imaginary parts of the complex reflection coefficients are obtained. The phase angle of the complex reflection coefficients in the complex plane is determined by the imaginary and real parts using a two-parameter arctangent function. The phase angles are arranged according to the frame number to obtain the phase angle sequence.

[0008] As a further aspect of the present invention: the process of calculating the signal disturbance index is as follows: In the phase angle sequence, the phase difference value is obtained by subtracting the previous phase angle from the current phase angle. All the phase difference values ​​are sorted in the order of the phase angle sequence to obtain the phase difference sequence. Set a sliding time window containing multiple consecutive frames, calculate the mean of multiple phase difference values ​​within the sliding time window, calculate the square mean of the difference between all phase difference values ​​within the sliding time window and the mean, and take the square root to obtain the signal disturbance index corresponding to the sliding time window. The signal disturbance index is updated once every time the sliding time window advances by one frame. Where the phase difference value does not belong to the interval Then by adding / subtracting The phase difference value is adjusted back to the range in this way. Inside.

[0009] As a further aspect of the present invention: the process of obtaining the baseline value is as follows: During initialization, during the period when the railway turnout throat area is not occupied by vehicles, the signal disturbance index corresponding to multiple sliding time windows is continuously acquired, and the mean and standard deviation of multiple signal disturbance indexes are calculated as the benchmark mean and benchmark standard deviation and stored. When performing anomaly detection for the first sliding time window, read the stored baseline mean and baseline standard deviation and perform anomaly detection at this time; After completing the anomaly detection for the first sliding time window, determine whether to update the benchmark mean and benchmark standard deviation. If yes, save the updated benchmark mean and benchmark standard deviation and delete the original benchmark mean and benchmark standard deviation. If no, save the original benchmark mean and benchmark standard deviation. When each subsequent sliding time window enters the anomaly detection phase, the stored baseline mean and baseline standard deviation are read, and the above anomaly detection and update process is executed.

[0010] As a further aspect of the present invention: the process of updating the benchmark mean and benchmark standard deviation is as follows: Obtain the current baseline mean A1 and baseline standard deviation A2, and calculate the baseline update threshold A3 = A1 + η1A2, where η1 is a preset first multiple; If the signal disturbance index of the current sliding time window is less than the baseline update threshold, and no intrusive disturbance marker is generated in the current sliding time window, the signal disturbance index is written into the sample set. When the number of signal disturbance indicators in the sample set exceeds the preset update number, the mean and standard deviation of the signal disturbance indicators in the sample set are calculated as the mean and standard deviation of the candidate baseline. The updated baseline mean A1' is obtained by weighting the candidate baseline mean and baseline mean A1 with preset weights, and the updated baseline standard deviation A2' is obtained by weighting the candidate baseline standard deviation and baseline standard deviation A2 with the same set of preset weights.

[0011] As a further aspect of the present invention: the process of outputting the turnout locking control signal is as follows: Obtain the current baseline mean B1 and baseline standard deviation B2, and calculate the discrimination threshold B3 = B1 + η2B2, where η2 is a preset second multiple and η2 > η1; If the signal disturbance index corresponding to the current sliding time window exceeds the discrimination threshold, and within the predetermined observation window, the number of signal disturbance indices exceeding the discrimination threshold is greater than the preset number threshold, then an intrusion disturbance marker is generated. After generating an intrusion disturbance marker, the main controller sends a high-level signal to the relay connected in series in the power supply circuit of the pneumatic solenoid directional valve through the digital output module. The relay coil is energized, driving the normally closed contact to open and cutting off the power supply circuit to the two solenoid coils of the dual-electric control three-position five-way neutral closed pneumatic solenoid directional valve. The valve core of the pneumatic solenoid directional valve returns to the neutral position under the action of the reset spring, closing the passage of the two air chambers of the cylinder and maintaining the switch in its current position.

[0012] A radar-sensing-based railway turnout safety control system includes: Acquisition module: Controls the millimeter-wave radar to transmit electromagnetic wave signals towards the throat area of ​​the railway turnout and receives the echo signals; Processing module: Processes the echo signal to generate digital difference frequency data, and then performs a fast Fourier transform on multiple consecutive frames of digital difference frequency data to obtain the range-velocity spectrum matrix; Optimization module: Select a reference cell in the range-velocity spectrum matrix, extract the phase angle of the reference cell frame by frame and establish a phase angle sequence, and calculate the signal disturbance index based on the phase angle sequence; Read the currently stored baseline value, which includes the baseline mean and baseline standard deviation, and determine the current discrimination threshold based on the baseline value; Control module: compares the signal disturbance index with the discrimination threshold, generates an intrusion disturbance flag when the discrimination condition is met, and outputs a turnout locking control signal based on the intrusion disturbance flag.

[0013] The beneficial effects of this invention compared to the prior art are as follows: This invention breaks away from the conventional approach of directly detecting strong reflected echoes from vehicles. Instead, it uses suspended conductive dust particles in the environment as the detection medium. By extracting and monitoring the phase perturbation complexity of the radar's fixed scattering reference unit, it achieves indirect detection of vehicle intrusion. When a metal vehicle enters the throat area of ​​a turnout, it disrupts the electromagnetic order of the dust scattering field, causing a sharp increase in the phase perturbation index, thereby triggering locking. This method does not rely on the severely attenuated vehicle target echo intensity, thus bypassing the electromagnetic shielding problem caused by conductive dust. The phase perturbation signal is highly sensitive and can be captured before the vehicle fully enters, providing valuable lead time for turnout locking. Simultaneously, the dynamic baseline self-learning mechanism continuously tracks the gradual changes in environmental noise and dust concentration, adaptively adjusting the discrimination threshold, significantly reducing the risk of false alarms and missed alarms, and greatly improving the reliability of turnout safety control in harsh industrial environments. Attached Figure Description

[0014] The invention will now be further described with reference to the accompanying drawings.

[0015] Figure 1 This is a flowchart illustrating a radar-sensing-based railway turnout safety control method according to the present invention. Figure 2 This is a schematic diagram of the process for obtaining baseline values ​​according to the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Please see Figures 1-2 As shown, this invention is a radar-sensing-based railway turnout safety control method, comprising the following steps: Control the millimeter-wave radar to transmit electromagnetic wave signals toward the throat area of ​​the railway turnout and receive the echo signals; The echo signal is processed to generate digital difference frequency data, and then a fast Fourier transform is performed on multiple consecutive frames of digital difference frequency data to obtain the range-velocity spectrum matrix. In a preferred embodiment of the present invention, the process of obtaining the range-velocity spectrum matrix is ​​as follows: The millimeter-wave radar is installed in a location capable of covering the throat area of ​​a railway turnout, with the transmission direction facing the direction in which vehicles enter the throat area. During detection, a voltage-controlled oscillator continuously generates a linear frequency modulated (LFM) continuous wave signal according to a preset frequency modulation slope. This LFM continuous wave signal is used as the transmission signal and is split into a transmit branch signal and a local oscillator branch signal by a power divider. The transmit branch signal is sent to the transmitting antenna, which then transmits electromagnetic wave signals towards the throat area of ​​the railway turnout. The local oscillator branch signal is sent to the receiving processing link for mixing with the echo signal received by the receiving antenna. After the electromagnetic wave signal reaches the throat area of ​​the railway turnout, it is scattered by conductive dust particles, track structures, fixed objects beside the track, and objects entering the area. The scattered electromagnetic waves are received by the receiving antenna, forming an echo signal.

[0018] The echo signal output from the receiving antenna is first amplified by low-noise processing to obtain the amplified echo signal. Then, the amplified echo signal and the local oscillator branch signal are simultaneously fed into a mixer for mixing, resulting in the difference frequency component between the two signals. The analog signal corresponding to this difference frequency component is denoted as the difference frequency signal. Since the frequency of the linear frequency modulated continuous wave signal varies with time, scatterers at different distances correspond to different echo propagation delays. These different propagation delays result in different difference frequency frequencies after mixing; therefore, the difference frequency signal contains information about the scatterer distance.

[0019] After the difference frequency signal enters the filter, the DC component generated by mixing and high-frequency noise exceeding the effective difference frequency range are filtered out, resulting in a filtered difference frequency signal. The analog-to-digital converter then samples the filtered difference frequency signal at a preset sampling frequency to obtain digital sample values. Multiple digital sample values ​​arranged in sampling time order within the same frequency modulation cycle are recorded as the digital difference frequency sequence corresponding to that frequency modulation cycle. The digital difference frequency sequences corresponding to multiple consecutive frequency modulation cycles (the specific number can be manually set) are combined to form a frame of digital difference frequency data.

[0020] For a frame of digital difference frequency data, a one-dimensional fast Fourier transform is first performed on the digital difference frequency sequence corresponding to each frequency modulation cycle to obtain multiple complex frequency components arranged according to frequency indices. Each frequency indices correspond to a difference frequency, and each difference frequency corresponds to an echo propagation delay. Based on the modulation slope of the linear frequency modulated continuous wave signal, the difference frequency is converted into an echo propagation delay. Then, based on the electromagnetic wave propagation speed and the round-trip propagation relationship, the echo propagation delay is converted into a scattering distance. The scattering distance corresponding to the frequency indices is used as the center distance of a range cell, and the complex frequency component at the frequency indices is used as the complex reflection coefficient of the range cell, thereby obtaining the range spectrum corresponding to the frequency modulation cycle.

[0021] Multiple range spectra within the same frame of digital difference frequency data are arranged in chronological order according to the frequency modulation cycle to form a range spectrum time matrix. The rows of this matrix correspond to different range cells, and the columns correspond to different frequency modulation cycles. Multiple complex reflection coefficients within the same row represent the changes in the same range cell over several consecutive frequency modulation cycles. For example, if a frame of digital difference frequency data includes four frequency modulation cycles, and each cycle yields five range cells, the range spectrum time matrix will contain five rows and four columns. Each row represents the sequence of complex reflection coefficients for a range cell across the four frequency modulation cycles. Subsequently, a one-dimensional fast Fourier transform is performed on each row along the column direction of the range spectrum time matrix, converting the changes in the frequency modulation cycle sequence into slow-time frequency components. This transforms the frequency modulation cycle columns in the original matrix into velocity channel columns, thus forming a range-velocity spectrum matrix. The rows of this matrix still correspond to range cells, and the columns correspond to velocity channels. Each element in the matrix represents the complex reflection coefficient for the corresponding range cell and velocity channel.

[0022] It should be noted that each element in the range-velocity spectrum matrix represents the complex reflection coefficient for the corresponding range cell and velocity channel. Millimeter-wave echoes can be represented as complex signals at the receiving end. The modulus of the complex number represents the strength of the scattered component, and the phase of the complex number represents the phase relationship between the scattered component and the local oscillator signal. Range-oriented Fast Fourier Transform (FFT) separates the range components corresponding to different difference frequencies, while slow-time FFT separates the phase change frequencies corresponding to different radial velocities within the same range cell. After these processing steps, each element in the matrix corresponds to the complex superposition result of the scattered components within a range and a velocity range, and this complex result contains both amplitude and phase information.

[0023] The formation of the range-velocity spectrum matrix relies on the principles of linear frequency modulated (LFM) continuous wave ranging and Doppler velocities. The LFM radar's transmission frequency changes continuously with time. Scatterers at different distances produce different echo propagation delays, which, after mixing, manifest as different difference frequencies, corresponding to the scatterer distances. As the scatterer moves along the radar's line-of-sight, the echo phase changes continuously between continuous frequency modulation cycles. This phase change corresponds to a slow-time frequency shift, which in turn corresponds to the scatterer's radial velocity. By first distinguishing the range component using a range-to-fast Fourier transform (FFT), and then distinguishing the velocity component using a slow-time FFT, the scattering information mixed in the echoes can be organized into a data matrix with both range and velocity dimensions.

[0024] Performing a one-dimensional Fast Fourier Transform (FFT) on the digital difference frequency sequence corresponding to each frequency modulation (FM) cycle can separate the multiple difference frequency components mixed within that FM cycle. The difference frequency signal within a FM cycle is formed by the superposition of echoes from multiple scatterers at different distances. Different scatterers correspond to different propagation delays, and different propagation delays correspond to different difference frequencies. The Fourier Transform decomposes the time-varying digital difference frequency sequence into multiple frequency components. The complex result at each frequency index represents the amplitude and phase of the digital difference frequency sequence at the corresponding difference frequency. In linear frequency modulated continuous wave (LFM) radar, there is a correspondence between the difference frequency and the scattering distance; therefore, these complex frequency components arranged by frequency index can be further mapped to different range cells.

[0025] Using the complex frequency components at the frequency indices as the complex reflection coefficients of the range cell preserves the amplitude and phase information of the scattered echo within that cell. Each frequency indice corresponds to a difference frequency, which in turn corresponds to an echo propagation delay, and this propagation delay corresponds to a scattering distance range. The complex frequency components output by the Fast Fourier Transform represent the complex response of the echo within that scattering distance range at the corresponding difference frequency; their real and imaginary parts together determine the amplitude and phase. Since the phase angle of the reference cell needs to be extracted subsequently, the range spectrum must retain not only the amplitude values ​​but also the complex results.

[0026] A reference cell is selected in the range-velocity spectrum matrix, the phase angle of the reference cell is extracted frame by frame and a phase angle sequence is established, and the signal disturbance index is calculated based on the phase angle sequence. In a preferred embodiment of the present invention, the process of obtaining the phase angle sequence is as follows: First, read the velocity dimension arrangement order of the range-velocity spectrum matrix to determine the velocity channel with zero radial velocity, and record the column number of the column containing this velocity channel as column index A. Then, within the range of two columns to the left and two columns to the right of column index A, read the amplitude value of the corresponding matrix element for each range cell. If a range cell has an amplitude value higher than the noise threshold in multiple consecutive frames of the range-velocity spectrum matrix, then this range cell is recorded as a fixed scattering range cell. The noise threshold can be pre-stored, or it can be determined by reading the amplitude value of the blank range region when there are no vehicles occupying the area, and then based on the background level of the amplitude value.

[0027] The fixed scattering distance units are arranged from nearest to farthest according to their distance unit numbers. If the numbers of adjacent fixed scattering distance units are consecutive, these adjacent fixed scattering distance units are combined into the same fixed scattering distance segment. Then, the center distance corresponding to each fixed scattering distance unit is read, and the fixed scattering distance units whose center distance is within the range of the safety protection zone are selected as candidate units.

[0028] For each candidate cell, its amplitude value within two columns to the left and right of column index A in the range-velocity spectrum matrix of consecutive frames is read, and the average amplitude is calculated. The candidate cell with the largest average amplitude is selected as the reference cell. After determining the reference cell, in the range-velocity spectrum matrix of each frame, the matrix element corresponding to the reference cell and the zero-velocity channel is read, and this matrix element is used as the complex reflection coefficient of the reference cell in that frame. The real and imaginary parts of the complex reflection coefficient are obtained, and the imaginary and real parts are processed using a two-parameter arctangent function to obtain the principal phase angle of the complex reflection coefficient in the complex plane. This principal phase angle is recorded as the phase angle corresponding to that frame. After obtaining multiple phase angles in consecutive frames, the multiple phase angles are arranged in order of frame number to obtain a phase angle sequence.

[0029] The zero-velocity channel corresponds to a fixed scattering portion where the radial velocity is zero or close to zero. Distance cells within the two columns to the left and right of column index A that consistently exceed the noise threshold indicate a stable scattering echo at that distance location during continuous observation. Adjacent distance cells correspond to spatially adjacent distance ranges. Combining adjacent distance cells into a fixed scattering distance segment allows for the representation of a continuous scattering region formed by the track structure, fixed objects along the trackside, and conductive dust under low-disturbance conditions. The safety protection zone is the distance range within which a locking judgment needs to be made before the vehicle reaches the critical position of the turnout throat. Selecting distance cells within the safety protection zone from the fixed scattering distance segment ensures that subsequent phase observations fall into the spatially required position for pre-judgment. The average amplitude reflects the strength of the scattering echo of the distance cell in the multi-frame distance-velocity spectrum matrix. Distance cells with larger amplitudes have clearer real and imaginary parts of the complex reflection coefficient, and the complex phase is less dominated by random noise. Using the candidate cell with the largest average amplitude as the reference cell provides a phase observation object located within the safety protection zone, with a clear spatial position and stable scattering echo, providing a fixed reference for subsequent phase changes caused by vehicle disturbance of conductive dust.

[0030] The reference cell in the range-velocity spectrum matrix corresponds to a specific range position and velocity channel. The matrix elements are complex reflection coefficients, derived from the complex projection of the echo signal onto the corresponding range frequency and slow time frequency. The real and imaginary parts of the complex signal correspond to the in-phase and quadrature components, respectively, together representing the phase state of the scattered echo relative to the local oscillator signal. The two-parameter arctangent function determines the principal phase angle based on the coordinates of the imaginary and real parts in the complex plane, distinguishing the phase direction in different quadrants and obtaining the phase angle of the reference cell in that frame. Multiple consecutive frames of the range-velocity spectrum matrix reflect the echo changes of the same reference cell over time. Extracting the phase angle of the same reference cell frame by frame and arranging them according to frame number forms a phase angle sequence describing the phase state of a fixed scattering reference point within the safety protection area as a function of time. When a vehicle enters the safety protection area, it causes changes in the position and motion of conductive dust particles, altering the propagation paths and coherent superposition relationships of multiple scattering components within the reference cell. This phase angle sequence provides continuous data for subsequent calculations of phase differences between adjacent frames and signal disturbance indices.

[0031] In a preferred embodiment of the present invention, the process of calculating the signal disturbance index is as follows: In the phase angle sequence, the phase difference value is obtained by subtracting the previous phase angle from the current phase angle. All the phase difference values ​​are sorted in the order of the phase angle sequence to obtain the phase difference sequence. Set a sliding time window containing multiple consecutive frames, calculate the mean of multiple phase difference values ​​within the sliding time window, calculate the square mean of the difference between all phase difference values ​​within the sliding time window and the mean, and take the square root to obtain the signal disturbance index corresponding to the sliding time window. The signal disturbance index is updated once every time the sliding time window advances by one frame. If the phase difference value does not belong to the interval [-π, π], then the phase difference value is adjusted back to the interval [-π, π] by adding / subtracting 2π.

[0032] It should be noted that in this invention, the “frame” in each forward frame of the sliding time window is consistent with the frame of the aforementioned digital difference frequency data. Both refer to a frame of digital difference frequency data composed of digital difference frequency sequences corresponding to multiple consecutive frequency modulation cycles, rather than a single frequency modulation cycle.

[0033] In this invention, one frequency modulation period corresponds to one digital difference frequency sequence. One digital difference frequency sequence consists of multiple digital sampled values ​​obtained by continuous sampling within the frequency modulation period. A one-dimensional fast Fourier transform is performed on a digital difference frequency sequence to obtain a range spectrum corresponding to the frequency modulation period. A range spectrum includes multiple range cells and the complex reflection coefficients corresponding to each range cell. Multiple consecutive frequency modulation periods correspond to multiple digital difference frequency sequences. The multiple digital difference frequency sequences corresponding to the multiple consecutive frequency modulation periods are collectively recorded as one frame of digital difference frequency data.

[0034] One-dimensional fast Fourier transform is performed on multiple digital difference frequency sequences within a frame of digital difference frequency data to obtain multiple range spectra. These range spectra are then arranged into a range spectrum time matrix according to the order of the frequency modulation cycles. A one-dimensional fast Fourier transform is then performed along the slow time dimension of the range spectrum time matrix, which represents the order of the frequency modulation cycles, to obtain a range velocity spectrum matrix corresponding to that frame of digital difference frequency data. Therefore, there is a one-to-one correspondence between a frame of digital difference frequency data and a range velocity spectrum matrix.

[0035] Multiple consecutive frames of digital difference frequency data are processed sequentially according to their frame numbers to obtain multiple consecutive range-velocity spectrum matrices. In each range-velocity spectrum matrix, a reference cell corresponds to a specific range cell and a specific velocity channel, and the reference cell has a complex reflection coefficient. The complex reflection coefficient of the reference cell is read from the range-velocity spectrum matrix corresponding to each frame, and a phase angle corresponding to that frame is obtained based on the real and imaginary parts of the complex reflection coefficient. Multiple consecutive frames of digital difference frequency data are processed sequentially to obtain multiple phase angles, which are then arranged according to their frame numbers to obtain a phase angle sequence. The phase angles corresponding to two adjacent frames are subtracted to obtain a phase difference value. Multiple consecutive phase difference values ​​form a phase difference sequence. Each time the sliding time window advances by one frame of digital difference frequency data, a new phase angle and a phase difference value are added, and the signal disturbance index corresponding to the sliding time window is updated accordingly.

[0036] Read the currently stored baseline value, which includes the baseline mean and baseline standard deviation, and determine the current discrimination threshold based on the baseline value; In a preferred embodiment of the present invention, the process of obtaining the baseline value is as follows: During the initialization phase, it is first confirmed that the railway turnout throat area is free of vehicles. In this state, multiple frames of digital difference frequency data are continuously collected. Following the aforementioned calculation process for phase angle sequence and phase difference sequence, signal disturbance indices corresponding to multiple sliding time windows are obtained sequentially. The average value of these multiple signal disturbance indices is calculated by summing all the indices and dividing by the number of indices to obtain the baseline mean. Then, the difference between each signal disturbance indice and the baseline mean is calculated, and the square of each difference is taken to obtain the average value. The square root of this average value is then used to obtain the baseline standard deviation. The baseline mean and baseline standard deviation are written into the storage area as the baseline values ​​read during anomaly detection in the first sliding time window.

[0037] When the first sliding time window enters anomaly detection, the currently stored baseline mean and baseline standard deviation are read. The read baseline mean is used as the current baseline mean, and the read baseline standard deviation is used as the current baseline standard deviation. Before completing the anomaly detection for this sliding time window, the currently stored baseline mean and baseline standard deviation are not rewritten. After completing the anomaly detection for the first sliding time window, the current baseline mean A1 and baseline standard deviation A2 are read, and the baseline update threshold A3 is calculated. If the signal disturbance index corresponding to the current sliding time window is less than the baseline update threshold A3, and no intrusion disturbance marker is generated in the current sliding time window, then the signal disturbance index is written into the sample set. If the signal disturbance index corresponding to the current sliding time window is not less than the baseline update threshold A3, or if an intrusion disturbance marker has been generated in the current sliding time window, then the signal disturbance index is not written into the sample set.

[0038] The sample set is used to store signal disturbance indicators under low-disturbance conditions. When the number of signal disturbance indicators stored in the sample set exceeds the preset update count, the average value of all signal disturbance indicators in the sample set is calculated to obtain the candidate baseline mean, and the standard deviation of all signal disturbance indicators in the sample set is calculated to obtain the candidate baseline standard deviation. Subsequently, the currently stored baseline mean A1 and baseline standard deviation A2 are read, and the candidate baseline mean and baseline mean A1 are weighted and summed using preset weights to obtain the updated baseline mean A1'; the candidate baseline standard deviation and baseline standard deviation A2 are weighted and summed using the same set of preset weights to obtain the updated baseline standard deviation A2'. The updated baseline mean A1' and updated baseline standard deviation A2' are written to the storage area, and the original baseline mean and baseline standard deviation are deleted so that the updated baseline value is read when the subsequent sliding time window enters anomaly detection. If the number of signal disturbance indicators in the sample set does not exceed the preset update count, the original baseline mean and baseline standard deviation are continued to be stored. Each subsequent sliding time window is executed in the following order: first, read the currently stored benchmark mean and benchmark standard deviation for anomaly detection; then, after anomaly detection is completed, determine whether to update the benchmark mean and benchmark standard deviation.

[0039] If the signal disturbance index corresponding to the current sliding time window is less than the baseline update threshold, and no intrusion disturbance marker is generated within the current sliding time window, then the signal disturbance index is within the low disturbance range defined by the current baseline mean and the current baseline standard deviation, and has not been identified as a disturbance state corresponding to vehicle intrusion. In statistics, the mean and standard deviation are used to describe the central location and dispersion of samples of the same class. The dynamic baseline needs to be updated by samples in a non-intrusive or low-disturbance state. Intrusive disturbance samples will cause the baseline to shift towards an abnormal state. The baseline update threshold is determined by the current baseline mean and the current baseline standard deviation. When the current signal disturbance index is lower than this threshold, it indicates that the sample still belongs to the sample near the current background disturbance distribution; the absence of an intrusion disturbance marker further excludes the window where vehicle disturbance has already occurred. By writing the signal disturbance index into the sample set, the sample set can preserve data consistent with the background dust scattering fluctuations. When the candidate baseline mean and candidate baseline standard deviation are subsequently calculated using this sample set, the obtained candidate baseline still represents the disturbance level under non-intrusion conditions. This allows the subsequent discrimination threshold to maintain its adaptability to background disturbances and provides a stable reference for disturbance discrimination when a vehicle enters the safety protection area.

[0040] The same set of weights means that when updating the baseline mean and baseline standard deviation, the candidate statistics and the current statistics are assigned the same weights. For example, when weighting the candidate baseline mean and the current baseline mean, the candidate baseline mean corresponds to the candidate item weight, and the current baseline mean corresponds to the historical item weight; when weighting the candidate baseline standard deviation and the current baseline standard deviation, the same candidate item weight and the same historical item weight are also used. Weighted summation is a smoothing and fusion process in statistical updates. The candidate statistics reflect the distribution of recent low-perturbation samples, and the current statistics reflect the stored historical baseline state. The new baseline obtained after weighting both retains both the historical and recent states. The baseline mean represents the central location of the background perturbation indicators, and the baseline standard deviation represents the dispersion of the background perturbation indicators. Together, they form the statistical basis for the discrimination threshold.

[0041] The signal disturbance index is compared with the discrimination threshold. When the discrimination condition is met, an intrusion disturbance flag is generated. Based on the intrusion disturbance flag, the turnout locking control signal is output.

[0042] In a preferred embodiment of the present invention, the process of outputting the turnout locking control signal is as follows: After calculating the signal disturbance index in each sliding time window, the currently stored baseline mean B1 and baseline standard deviation B2 are read first, and a discrimination threshold B3 is calculated. The signal disturbance index corresponding to the current sliding time window is compared with the discrimination threshold B3. If the signal disturbance index does not exceed the discrimination threshold B3, no intrusion disturbance flag is generated, and the process proceeds to the next sliding time window. If the signal disturbance index exceeds the discrimination threshold B3, the sliding time window is recorded as an out-of-limit window, and the number of out-of-limit windows continues to be counted within a predetermined observation window. The observation window consists of multiple consecutive sliding time windows. For example, if the observation window contains 5 consecutive sliding time windows, the number of windows in which the signal disturbance index exceeds the discrimination threshold B3 is counted sequentially within these 5 sliding time windows. When the number of out-of-limit windows is greater than a preset threshold, an intrusion disturbance flag is generated. When the number of out-of-limit windows is less than the preset threshold, no intrusion disturbance flag is generated, and the statistical results within the observation window continue to be updated with new sliding time windows.

[0043] After an intrusion disturbance marker is generated, a high-level signal is output from the digital output port of the main controller and transmitted to the control terminal of the relay coil connected in series in the power supply circuit of the pneumatic solenoid directional valve, energizing the relay coil. Once energized, the relay coil drives the normally closed contact connected in series in the power supply circuit of the pneumatic solenoid directional valve to open, cutting off the power supply circuit to the two solenoid coils of the dual-electrically controlled three-position five-way neutral-position enclosed pneumatic solenoid directional valve. After the two solenoid coils are de-energized, the valve core of the pneumatic solenoid directional valve returns to the neutral position under the action of the return spring. In the neutral position, the passages of the two air chambers of the cylinder are closed, preventing the air paths on both sides of the cylinder from continuing to output reversing driving force to the switch mechanism. The switch remains in the position where the intrusion disturbance marker was generated.

[0044] η1 is used to form the baseline update threshold, and η2 is used to form the intrusion detection threshold. Both are based on the statistical descriptions of the baseline mean and baseline standard deviation. The baseline mean represents the central tendency of the signal disturbance index under no-vehicle occupancy conditions, and the baseline standard deviation represents the natural fluctuation range of the signal disturbance index under these conditions. The larger the multiple, the greater the deviation of the threshold from the baseline mean. The baseline update process needs to receive samples that are close to the background fluctuation state, so η1 is taken as a small multiple to keep the data entering the sample set within a low disturbance range. The intrusion detection process needs to identify disturbance states that are more obvious than the background fluctuations, so η2 is taken as a multiple greater than η1, making the detection threshold higher than the baseline update threshold.

[0045] The overall concept of this invention is to select a reference unit with stable scattered echoes within the safety protection area in front of the throat area of ​​a railway turnout, and continuously extract the complex reflection coefficient phase angle of the reference unit. When a vehicle enters the safety protection area, the movement of the vehicle body will drive the surrounding airflow, causing changes in the position and motion state of suspended conductive dust particles such as iron powder and steel slag. After the conductive dust particles participate in millimeter-wave scattering, they will change the propagation path and coherent superposition relationship between multiple scattering components within the reference unit, increasing the amplitude and dispersion of the phase angle variation between adjacent frames. Therefore, using the standard deviation of the phase difference between adjacent frames as a signal disturbance index can characterize the fluctuation of the dust scattering state within the safety protection area relative to the unoccupied state. Then, using the baseline mean and baseline standard deviation formed under the unoccupied state, a discrimination threshold is determined. By comparing the current signal disturbance index with the discrimination threshold, an intrusion disturbance mark can be generated when the dust scattering state caused by vehicle movement deviates significantly from the background state, and a turnout locking control signal can be output accordingly.

[0046] A radar-sensing-based railway turnout safety control system includes: Acquisition module: Controls the millimeter-wave radar to transmit electromagnetic wave signals towards the throat area of ​​the railway turnout and receives the echo signals; Processing module: Processes the echo signal to generate digital difference frequency data, and then performs a fast Fourier transform on multiple consecutive frames of digital difference frequency data to obtain the range-velocity spectrum matrix; Optimization module: Select a reference cell in the range-velocity spectrum matrix, extract the phase angle of the reference cell frame by frame and establish a phase angle sequence, and calculate the signal disturbance index based on the phase angle sequence; Read the currently stored baseline value, which includes the baseline mean and baseline standard deviation, and determine the current discrimination threshold based on the baseline value; Control module: compares the signal disturbance index with the discrimination threshold, generates an intrusion disturbance flag when the discrimination condition is met, and outputs a turnout locking control signal based on the intrusion disturbance flag.

[0047] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the present invention should still fall within the scope of the present invention.

Claims

1. A railway turnout safety control method based on radar sensing, characterized in that, Includes the following steps: Control the millimeter-wave radar to transmit electromagnetic wave signals toward the throat area of ​​the railway turnout and receive the echo signals; The echo signal is processed to generate digital difference frequency data, and then a fast Fourier transform is performed on multiple consecutive frames of digital difference frequency data to obtain the range-velocity spectrum matrix. A reference cell is selected in the range-velocity spectrum matrix, the phase angle of the reference cell is extracted frame by frame and a phase angle sequence is established, and the signal disturbance index is calculated based on the phase angle sequence. Read the currently stored baseline value, which includes the baseline mean and baseline standard deviation, and determine the current discrimination threshold based on the baseline value; The signal disturbance index is compared with the discrimination threshold. When the discrimination condition is met, an intrusion disturbance flag is generated. Based on the intrusion disturbance flag, the turnout locking control signal is output.

2. The railway turnout safety control method based on radar sensing according to claim 1, characterized in that, The process of obtaining the range-velocity spectrum matrix is ​​as follows: The linear frequency modulated continuous wave signal generated by the voltage-controlled oscillator of the millimeter-wave radar is used as the transmission signal and is divided into a transmission branch and a local oscillator branch by a power divider; The echo signal received by the receiving antenna is amplified by low noise and then mixed with the local oscillator branch to obtain a difference frequency signal containing the distance information of the scatterer; After the difference frequency signal is filtered to remove the DC component and high-frequency noise, it is converted into a digital difference frequency sequence by an analog-to-digital converter. A one-dimensional fast Fourier transform is performed on the digital difference frequency sequence within each frequency modulation cycle to obtain the range spectrum. The range spectra corresponding to multiple consecutive frequency modulation cycles are arranged into a matrix in time order, and then a one-dimensional fast Fourier transform is performed along the slow time dimension to form a range velocity spectrum matrix, where the frequency shift in the slow time dimension corresponds to the radial velocity of the scatterer.

3. The railway turnout safety control method based on radar sensing according to claim 1, characterized in that, The process of obtaining the phase angle sequence is as follows: In the range-velocity spectrum matrix, determine the column index A of the zero velocity channel. Each range cell in the zero velocity channel corresponds to the part of the scatterer where the radial velocity is zero. Search for range cells whose amplitude value is continuously higher than the noise threshold within the range of the two columns to the left and right of column index A. The searched adjacent distance cells are combined into a fixed scattering distance segment. The distance cells located within the safety protection area are selected as candidate cells, and the distance cell with the largest average amplitude is selected as the reference cell from the candidate cells. The complex reflection coefficients of the reference unit are obtained frame by frame, and the real and imaginary parts of the complex reflection coefficients are obtained. The phase angle of the complex reflection coefficients in the complex plane is determined by the imaginary and real parts using a two-parameter arctangent function. The phase angles are arranged according to the frame number to obtain the phase angle sequence.

4. The railway turnout safety control method based on radar sensing according to claim 1, characterized in that, The process of calculating the signal disturbance index is as follows: In the phase angle sequence, the phase difference value is obtained by subtracting the previous phase angle from the current phase angle. All the phase difference values ​​are sorted in the order of the phase angle sequence to obtain the phase difference sequence. Set a sliding time window containing multiple consecutive frames, calculate the mean of multiple phase difference values ​​within the sliding time window, calculate the square mean of the difference between all phase difference values ​​within the sliding time window and the mean, and take the square root to obtain the signal disturbance index corresponding to the sliding time window. The signal disturbance index is updated once every time the sliding time window advances by one frame. Where the phase difference value does not belong to the interval Then by adding / subtracting The phase difference value is adjusted back to the range in this way. Inside.

5. The railway turnout safety control method based on radar sensing according to claim 1, characterized in that, The process of obtaining the baseline value is as follows: During initialization, during the period when the railway turnout throat area is not occupied by vehicles, the signal disturbance index corresponding to multiple sliding time windows is continuously acquired, and the mean and standard deviation of multiple signal disturbance indexes are calculated as the benchmark mean and benchmark standard deviation and stored. When performing anomaly detection for the first sliding time window, read the stored baseline mean and baseline standard deviation and perform anomaly detection at this time; After completing the anomaly detection for the first sliding time window, determine whether to update the benchmark mean and benchmark standard deviation. If yes, save the updated benchmark mean and benchmark standard deviation and delete the original benchmark mean and benchmark standard deviation. If no, save the original benchmark mean and benchmark standard deviation. When each subsequent sliding time window enters the anomaly detection phase, the stored baseline mean and baseline standard deviation are read, and the above anomaly detection and update process is executed.

6. The railway turnout safety control method based on radar sensing according to claim 5, characterized in that, The process of updating the benchmark mean and benchmark standard deviation is as follows: Obtain the current baseline mean A1 and baseline standard deviation A2, calculate the baseline update threshold A3 = A1 + η1A2, where η1 is a preset first multiple; if the signal disturbance index of the current sliding time window is less than the baseline update threshold, and no intrusion disturbance marker is generated in the current sliding time window, write the signal disturbance index into the sample set; When the number of signal disturbance indicators in the sample set exceeds the preset update number, the mean and standard deviation of the signal disturbance indicators in the sample set are calculated as the mean and standard deviation of the candidate baseline. The updated baseline mean A1' is obtained by weighting the candidate baseline mean and baseline mean A1 with preset weights, and the updated baseline standard deviation A2' is obtained by weighting the candidate baseline standard deviation and baseline standard deviation A2 with the same set of preset weights.

7. A railway turnout safety control method based on radar sensing according to claim 6, characterized in that, The process of outputting the turnout locking control signal is as follows: Obtain the current baseline mean B1 and baseline standard deviation B2, calculate the discrimination threshold B3 = B1 + η2B2, where η2 is a preset second multiple and η2 > η1; if the signal disturbance index corresponding to the current sliding time window exceeds the discrimination threshold, and within the predetermined observation window, the number of signal disturbance indices exceeding the discrimination threshold is greater than the preset number threshold, then an intrusion disturbance marker is generated. After generating an intrusion disturbance marker, the main controller sends a high-level signal to the relay connected in series in the power supply circuit of the pneumatic solenoid directional valve through the digital output module. The relay coil is energized, driving the normally closed contact to open and cutting off the power supply circuit to the two solenoid coils of the dual-electric control three-position five-way neutral closed pneumatic solenoid directional valve. The valve core of the pneumatic solenoid directional valve returns to the neutral position under the action of the reset spring, closing the passage of the two air chambers of the cylinder and maintaining the switch in its current position.

8. A railway turnout safety control system based on radar sensing, characterized in that, include: Acquisition module: Controls the millimeter-wave radar to transmit electromagnetic wave signals towards the throat area of ​​the railway turnout and receives the echo signals; Processing module: Processes the echo signal to generate digital difference frequency data, and then performs a fast Fourier transform on multiple consecutive frames of digital difference frequency data to obtain the range-velocity spectrum matrix; Optimization module: Select a reference cell in the range-velocity spectrum matrix, extract the phase angle of the reference cell frame by frame and establish a phase angle sequence, and calculate the signal disturbance index based on the phase angle sequence; Read the currently stored baseline value, which includes the baseline mean and baseline standard deviation, and determine the current discrimination threshold based on the baseline value; Control module: compares the signal disturbance index with the discrimination threshold, generates an intrusion disturbance flag when the discrimination condition is met, and outputs a turnout locking control signal based on the intrusion disturbance flag.

Images